A texture analysis approach of selecting an optimum grid size

The complexity and computational cost of atmospheric flow and dispersion modeling depends to a large extent on the spatial resolution of the initial model setup. The choice of model grid resolution helps to determine the overall accuracy of the model results. In some cases, such as atmospheric modeling in complex terrain or dispersion modeling in areas with high emission gradients, the requirement for adequate horizontal resolution can have a critical impact on the quality of the final simulation. To determine the optimum spatial resolution to adequately model a particular problem, however, very little quantitative guidance is currently available.

The objective of this study is to develop a texture based analysis approach to quantitatively estimate the differences in the spatial distribution of model derived output fields. Texture analysis measures used to estimate contrast, local homogeneity, and level of randomness in the spatial distribution of model results will be described.

The technique will be applied to the simulated wind, temperature and moisture patterns from a series of MM5 simulations at variety of horizontal resolutions to access the adequacy of each resolution in capturing the essential details of the modeled atmospheric state. MM5 is a nested non-hydrostatic atmospheric modeling system which has been applied at many different horizontal scales. In this study, MM5 version 2.10 will be run at 108 km, 36 km, 12 km and 4 km in a one-way interactive nesting configuration for the period July 11-15, 1995.